Fig. 4

Clustered HMRs are enriched for active regulatory elements compared to unclustered HMRs. A Boxplot of ATAC-STARR-seq regulatory element overlap by clustered and unclustered HMRs. Overlap is measured at the unit of HMRs, and values depict fraction of total HMRs that contain a regulatory element. A Wilcoxon rank sum test was used to determine statistical significance. B Point and line graph of the proportion of HMRs near an expressed gene at different TSS distances. HMRs are grouped by HMR clusters that contain a cell specific HMR and unclustered cell specific HMRs. Denominators for the HMR clusters and unclustered HMR groups are 444 and 1621, respectively. Counts below the graph represent the cumulative amount of genes below each threshold per HMR group. p-values are derived from a z-test of proportions to test the fraction of HMRs represented by HMR-single nearest neighbor gene pairs below each threshold distance. C Boxplot of TPM values (derived from GM12878 cell line data) of nearest neighbor RefSeq protein-coding genes to clustered and unclustered HMRs. Two nearest neighbor genes (with TPM > 0) per HMR were filtered for TAD boundary crossing. Statistical significance was measured by a Wilcoxon rank sum test in R using the wilcox.test() function. D Multiple alignment of region around the CD27 locus showing methylation and HMR tracks across 6 cell types: H1 ESC, hematopoietic stem and progenitor cells, macrophage, neutrophil, B cell, and T cell. Methylation tracks are represented by orange vertical bars showing methylation value per CpG site. HMRs are shown by dark blue horizontal bars. Below the multiple alignment, Hi-C interaction score data is represented by heatmap triangles representing interaction matrices for GM12878s and H9 ESC cells. Values for the Hi-C data are derived from .hic interaction matrix files. The plotgardener R package was used to generate the genome browser snapshot [29]